Saturday, July 23, 2016

Bemusing aero equipment choices at the Tour de France

2016 Tour de France. Stage 18 ITT Megève. GC contenders giving away time with bike set up choices. Why?

Here's the course profile:

Here's a table with the aero choices made by the 20 fastest riders on the day. As far as I can tell all rode using a skin suit (although some of the suits were not exactly a good aero fit).

This suggests all these riders recognised that aerodynamics still mattered, but not enough that riders thought it worth using some other basic aero kit. Perhaps they felt there was too much of a weight penalty (there's not BTW). Or they did not feel good climbing on a TT bike, or were concerned with the descent? Lack of preparation is my take.

For reference, I used photos from the various websites to work out who used what. For front wheel and helmets, there might be a little debate as to it fits the category of aero or not. Needed to be a full aero TT helmet to count and what looked like low-ish profile wheels went in the "No" category. Always happy to amend if people spot errors.

Richie Porte, with not even an aero front wheel, let alone an aero helmet:

Fabio Aru, not much better:

Contrast with the stage winner Chris Froome who used all the aero aids at his disposal:


Average speeds for the top 20 ranged from 31.5km/h to 33.2km/h. Aerodynamics still matters quite a bit at such speeds. So why not take advantage of it?

Yes it was hilly but the lack of aero equipment choices for a TT even at these speeds does rather bemuse me. Weight penalty of helmets and wheels is negligible and any small benefits are outweighed by aero losses.

More discussion later. Perhaps.

Read More......

Sunday, December 13, 2015

FTP variability (and doping)

In one of the five hundred and twenty five thousand online forum threads about why Chris Froome is or is not a doper, one of the questions raised was about whether a coach could detect if an athlete was on the juice based on their performance (power) data.

That led to a comment about typical changes in a rider's power over the course of a season.

As to the question of a coach's ability to detect doping from performance, performance changes are multifactoral and so that makes it nigh on impossible.

It's relatively easy to measure the performance change (power meters enable that), far more difficult to parse out the specific reasons why it occurs.

Now of course one can wonder if you have known an athlete for a long time and know their training and performance history and have a reasonable understanding of their potential. If they find a sudden large boost when nothing else in particular has changed, well you might naturally begin to wonder.

Consider that I have seen athletes attain Functional Threshold Power improvement of between 5% and 100% in 6 months of training and you can immediately see the problem, especially given doping provides performance advantages well within the range of those attainable by completely legitimate means.

Better training, better diet, better sleep, better psychology, better aero, better planning and support, better race skills and race craft, better equipment and tools, and of course, doping. These are not mutually exclusive means to improve performance.

This is the problem e.g. that makes up much of the discussion about Froome or others. Lot's of Clinic focus on his "transformation". The problem is that there are plenty of legitimate as well as illegitimate means by which such performance changes can be explained.

Balance that with the fact that in the past 30 years half the riders standing on the podium for the major Euro pro races and top 20 in GTs are known to be dopers (let alone the ones that slipped though the net). Objective assessment therefore needs to consider all such possibilities.

However that still doesn't mean one can immediately infer from performance data or even physiological testing data such as lactate threshold or VO2max the reasons for one's performance, or more to the point, their change in performance.

I think the only way an ethical coach is likely to spot or suspect doping is if they are in frequent eye ball contact with the athlete, and it's not so much going to be from their on-bike performance, but rather from observing off-the-bike behaviour.

As much as coaches might like to be in frequent eye ball contact so they can do a better job, coaches are often not in such frequent close quarters with their clients. Riders travel and coach can't be with all their clients all the time. The exception are squad/institute coaches that interact multiple times per week and travel with their athletes that typically attend the same races.

More usually the contact is via phone/skype/chat/email and other social and electronic media style interactions, as well as the athlete's diary notes that accompany their power meter files. For the most part this works pretty well (athlete results demonstrate that to us all the time) but of course there are some things for which seeing the athlete is preferable and some personalities that require more eye ball contact than others.

Anyway, on one of the forums I made a comment about the typical variability in FTP for an active racing cyclist. An often quoted value is about 10% variance from out of form/off season to peak fitness. That was questioned as being quite a large variance. I really had nothing other than my years of coaching and personal experience to suggest whether or not this was realistic.

So I thought about attempting to answer the question with some data.

Fire up WKO4 and create a report using the following expression:

max(ftp(meanmax(power),90)) / min(ftp(meanmax(power),90))

and apply it to ranges covering entire years of data (with power data for >>90% of rides).

That expression calculates the modelled FTP for the date range selected, locates the maximum and minimum values for FTP that are calculated during that date range, and calculates the ratio of the maximum to the minimum FTP.

I did that for a selection of 10 athletes over 2 seasons. These athletes are mostly competitive amateur through to elite level (but no full time pros), and have power data for >> 90% of their rides.

This is the summary:

What I find interesting is the variance as measured by the modelled FTP in WKO4 is larger than I would have expected.

Over 10 riders for 2 seasons each, we have an average maximum to minimum modelled FTP ratio of 1.23, meaning the peak modelled FTP for a season was, on average, 23% higher than the minimum modelled FTP for that same season.

Good luck trying to pick out one specific reason for performance changes when models are showing this sort of variance in FTP.

Do I think their FTP really varies that much? Well possibly not quite, but then with time I am seeing mFTP to be quite reliable indicator, provided the quality of input data is good. One erroneous power spike can mess with the power-duration data and mFTP value. Indeed when there are large changes in the modelled power-duration metrics, it's often due to input data error than anything else.

For reference I also provided an indication of their annual TSS (~27,000) and average CTL (~77 TSS/day) for this selection, just to show that theses are riders on average have quite decent training volume. I would not rely totally on those TSS values though, as they probably need an audit of the FTP history applied in WKO4 to generate them, so I consider them as just indicative for now.

I also looked at my own data for 2009 and 2010, and my annual mFTP variance was 15% each year, so a bit lower than the average reported above.

Now of course with all such things one needs to consider context, and quality of the input data. For now that's a study beyond what I have time for.

Read More......

Friday, December 04, 2015

Looking under Froome's hood

A little over two years ago I wrote about the relationship between four key underpinning physiological parameters that determine a rider's sustainable power output:

  • VO2max
  • Energy yield from aerobic metabolism
  • Efficiency
  • Fractional utilisation of VO2max at threshold

I don't propose to repeat myself, so go here to read that first if you'd like a more detailed explanation.

Data on some physiological testing by Chris Froome was released earlier today, so I thought I'd put a marker on one of the charts I posted in that earlier item to see where he sits.

I took the data from the cyclingnews article linked below:

In it the key 2015 data are listed as:

Weight: Test: 69.9kg, TdF: 67kg
VO2max: Test: 84.6ml/kg/min, TdF weight adjusted: 88.2ml/kg/min
Threshold power (20-40 min): 419W
W/kg: 5.98W/kg, TdF weight adjusted: 6.25W/kg

So given we are talking 20+ minute power, a fractional utilisation of 90% of VO2max for an elite athlete is not unreasonable, so here's that particular chart, and overlayed on that is a pink box defining the area covering a range of VO2max from 75ml/kg/min to 95ml/kg/min and gross efficiency range from 19% to 25%. You'd expect elite cyclists to be somewhere in that range.

Froome's estimated TdF VO2max and 20+ minute power/mass are then shown by the green dot:

What can we infer from this?

Not a lot really, other than the data are in line with what you would expect for a rider with the performances of a grand tour winner. Certainly the physiological values are in line with historical data on plausible physiological parameters for elite aerobic endurance athletes.

As far as informing on doping status, as with power meter data and climbing power estimates, it tells us SFA. In any case I doubt it will change anyone's opinion either way.

Edit: here is a link to the lab report:

Read More......

Tuesday, October 13, 2015

Kona power meter usage trends: 2009 to 2015

Update for 2015 based on the Lava Magazine bike count data. Previous posts links showing trend data up to 2013 and 2014 are here:

Without further ado, here are the numbers for 2009 through to to 2015 are (click on images to see larger versions):

In brief, 2015 continued the long term trend of an increase in use of power meters by Kona IM athletes, with a tick under half of all bikes now fitted with a power meter.

The two longest established brands, SRM and Powertap, have further fallen away in absolute numbers as well as total share dropping with Powertap suffering the biggest drop in usage, and while Quarq is still the most used meter, its absolute usage has reached a plateau and it is no longer as dominant a power meter brand for Kona IM athletes as it has been in the past few years. It will be interesting to see how Powertap fares in the years ahead with the introduction of their new pedal and chainring based meters.

The use of power meters is more evenly distributed across the various brands than in previous years, with no brand dominating share of usage on Kona IM athlete's bikes.

Newer power meter brands have increased their presence significantly, in particular Garmin Vector and especially Stages being the big movers.

Power2Max maintained their 2014 share of the power meter pie, while newer offerings from Rotor and Pioneer make up the smaller slices.

Thanks to Prof. Hendrik Speck of Hochschule Kaiserslautern University of Applied Sciences for picking up a couple of very small errors in the Polar power meter numbers I had listed for 2009 and 2013. I have updated the table and chart above. I also left the linked posts from previous year's summaries uncorrected so that a record of the small error remains.

Read More......

Monday, August 24, 2015

When your ride buddy becomes a real drag

A question that comes up from time to time when chatting about aerodynamics stuff is how much impact does another rider in close proximity have on your aerodynamics, or more correctly stated, does having another rider in close proximity change the power required for you to maintain your speed?

We are all familiar with the reduction in power required when riding behind another rider. This "drafting" benefit is substantial and anyone with a power meter can see the big reduction in power when they move from riding directly into the wind to riding behind another rider. Even if you don't have a power meter the difference is certainly large enough to notice the reduction in effort required.

But what about when your buddy is drafting behind you or rides beside you? Does this impact the power needed to maintain the same speed?

The short answer is: yes, both of them do.
But in what way and by how much?

The question as to whether a rider in front gains benefit from having a rider behind them has been researched before, and the consensus is that yes, they gain a small benefit. There is good reason for this slightly counter intuitive result and it's to do with the "bow wave" of air from the rider behind causing a change in the turbulent air flow behind the lead rider and reducing, by a small amount, the depth of the low pressure zone that exists behind the front rider.

This slight reduction in the fore to aft air pressure differential of the lead rider provides a small but measurable gain. This can be expressed as a reduction in apparent CdA, but since a rider's CdA doesn't really change if their position and equipment hasn't, then in reality it's a change in the forces acting on the rider, and as a result, the power demand at the same speed is slightly reduced when compared with having no rider in close proximity (or alternatively, a rider can travel slightly faster for the same power when they have a rider immediately behind them).

In 2010  Andy Coggan examined data from a 2007 track session ridden by his wife, in which she did efforts on the track both with and without having a rider drafting behind her. In Andy's assessment of the data he remarked "having a rider drafting closely behind them apparently lowered their CdA by 3.2%, i.e., from 0.198 to 0.192 m^2.".

The reduction in energy demand will be of a very similar amount to the reduction in apparent CdA. Assuming ~350W, a reduction from a CdA of 0.198 to 0.192 is equivalent to a reduction in power demand at the same speed of ~10W, or 2.8%. In this case the other rider was riding in pursuit set up, and were themselves very "aero" (an elite track pursuit rider).

So that's one example.

This phenomenon has also been reported in the published scientific literature, examples include:

Racing cyclist power requirements in the 4000-m individual and team pursuits, Medicine and Science in Sports and Exercise, v31, no.11, pp 1677-1685, 1999. J.P. Broker, C.R. Kyle and E.R. Burke.

where amongst their data they report that the lead rider requires 2-3% less power while riding on the front of a 4-man team than if riding solo at the same speed.

Another more recent study examined this using both computational fluid dynamics (CFD) simulations along with wind tunnel validation as described in this paper:
CFD simulations of the aerodynamic drag of two drafting cyclists, Computers & Fluids Volume 71, 30 January 2013, Pages 435–445,. Bert Blocken, Thijs Defraeyeb, Erwin Koninckxc, Jan Carmelietd, Peter Hespelf

In this paper they report the lead rider of two riders riding in single file receives a reduction in energy demand of 2.6% while riding in the time trial position.

Above are three examples of data from a similar situation, with reported reductions in energy (power) demand to ride at the same speed ranging between 2% to 3% for the lead rider compared with riding solo.

There's another paper that reports a 5% advantage for the lead rider of team time trial, although I'm not able to see more than the abstract:

Aerodynamics of a cycling team in a time trial: does the cyclist at the front benefit?; European Journal of Physics, Volume 30 Number 6, 2009; A Íñiguez-de-la Torre and J Íñiguez

Edit: I've now read the paper and it used two dimensional CFD analysis on ellipses as a simple model simulation of multiple riders in a line and is indicative of the principles involved.

I've had the resources to test this for some time but I've hadn't got around to doing the experiment, mainly because exclusive use of track time costs money and I'm focussed on working with clients on answering more important aerodynamics questions for them than doing experiments just for the fun of it.

But today I had the opportunity to do just such an experiment.

I was doing aerodynamics testing as part of a story being written about a woman masters rider preparing for the UCI World Masters track cycling championships being held in Manchester later this year. Cycling NSW kindly offered and arranged for the track time to make this possible, and a client of mine, Rod Wagner, loaned a special power meter to enable the testing on the rider's track bike, while I offered my time for the aero work.

We'd reached the end of our allotted track time, but as luck would have it no one else was ready to ride on the track, so we had some spare time for the experiment, and willing participants.

I won't comment on the primary aero testing session as that's for another to write about for later publication in magazine and online, but I'll expand on the impromptu experiment.

The method of measurement

With the Alphamantis Track Aero System, I record and monitor in real time a rider's aerodynamics as they circulate around the indoor velodrome. Testing is done indoors as this removes the wind variable and provides for a well controlled environment. The system enables us to monitor speed and velocity and along with other key inputs such as air density, track geometry data, centre of mass height, rider mass and rolling resistance variables, the Coefficient of drag x Frontal area (CdA) is also plotted in real time and lap by lap a picture of a rider's aerodynamics is revealed.

I've briefly explained this system before in this post, which also has a video demo. You can also read more on the Alphamantis site linked above.

The experiment

Normally this testing is done with a rider riding solo on the track but for this experiment I asked her coach, another world level master's rider, to join in. His task was to ride in various positions relative to the test rider (who would continuously circulate around the track at approximately 40km/h) while her coach would change his relative position on the track every 4-6 laps as follows and on my instruction, he would:

- ride in front of the test rider to test the level of drafting assistance, then
- ride next to, and on the outside of the test rider, then
- ride immediately behind the test rider, then
- drop off entirely and stop riding, so that we could obtain data from the test rider circulating solo.

This test process was repeated a second time during the long test run to validate the results from the first run.

For reference, the test rider is a slim 60kg female approximately 172cm tall, and the coach weighs approximately 80kg and is ~185cm tall. The test rider was using a track bike with pursuit bars, while the other rider was using a track bike in regular mass start set up.

The system is really reporting the impact on apparent CdA. It's a quick way to measure how beneficial or detrimental having the other rider near you is, and the measurements are not overly sensitive to the changes in speed during the run (this is the nice thing about the process).

The results

Here's a table summarising the results of all the data runs. Click on images to see larger versions.

In the case of the support rider riding behind the test rider, the test rider gained a benefit of a reduction in apparent CdA of around 0.008m^2, or about 3.8%. Note (i) the error range and (ii) the support rider was riding in a more upright mass start position (and consequently has a larger "bow wave") and is somewhat larger than the test rider.

Also shown are the results of the traditional drafting, being a reduction in apparent CdA to nearly half of the solo value, and interestingly, the apparent CdA increase of ~ 0.013m^2, or nearly 6% when the other rider was riding alongside the test rider.

Since apparent CdA differences are a little harder to understand, I've flipped the data around to show, at a normalised velocity of 40km/h, what the power demand for the solo rider would be for each scenario:

The table below summarises the chart data, and also shows the difference in power compared with riding solo:

Compared with riding solo, the test rider gains a ~7W (3%) benefit from having her ride buddy directly behind her; a 76W (39%) benefit from drafting behind her ride buddy; and a 10W (5%)  penalty when her ride buddy is riding alongside.

So in this experiment, I found a 3% energy demand benefit from having a trailing rider, and that's right in line with (but at the top of) the range found by the other reported data referenced earlier.

This result of a 10W penalty when riding alongside another rider is more novel, although it doesn't surprise me it may be news to some.

It is something to ponder when riding in team formation events, especially when the lead rider pulls aside to make their way to the back of the line of riders. They and their team are better off (at least in low yaw conditions) if the rider pulls over and moves well away from their companions until they are near the back and can return to be in the draft of the other riders. 10W is nothing to sneeze at.


So it would seem that if you wish to ride alongside your ride buddy, it might cost you ~10W give or take. If speed is of the essence, then ride in single file, you'll both go quicker that way.

Read More......

Sunday, July 26, 2015

Alpe d'Huez: TDF Fastest Ascent Times 1982-2015

Update of the Alpe d'Huez climbing times and speed chart previously posted here and here. Read those previous posts for discussion of context.

Edit (28 July 2015): since posting this two days ago, I was alerted to some updates made to the 1991 ascent times. Two sources did work with archive video to better verify these times, the net result being an addition of 41 seconds to each of the 1991 ascent times.

Thanks to for the posting the data.

This chart shows the average speed of the five fastest ascents up the Alpe d'Huez climb for each year the Tour de France included this climb, with the exception being the times from the 1980s which are the average speeds for fewer riders (as data on five fastest ascents in those years is not available to me).

As a reminder, I chose to average the 5 fastest ascent times for a couple of reasons:
- it reduces the individual noise in the data for year by year comparisons
- the 5 fastest were most likely to have been giving it close to maximal effort and would be representative of the quality at pointy end of the field
- the available historical data I have on ascent times doesn't permit increasing that sample size all that much in any case.

 Here's the data in table format, along with some extra context information. I've also ranked the average ascent speeds of the 5 fastest for each of the 13 occasions during 1991-2015 that Alpe d'Huez was climbed. I left out ranking 1980s ascents as I don't have times for all 5 fastest riders for those years (IOW the actual average speed of 5 fastest would be lower).

As we can see, 2015 ranks as the 8th fastest TdF ascent over that period, when based on the 5 fastest ascents each year.

Here's the same table but with weather conditions for the airport nearest to Boug d'Oisans listed from 3pm to 5pm on the day of the race. I was only able to source data back to 1997. If anyone knows of an online almanac of weather data for near Bourg d'Oisans for years prior to 1996, please let me know.

Weather data source:
Note the variability in temperature from year to year, and importantly the prevailing wind direction and speed. 

Now how such prevailing wind actually plays out on the slopes of the Alpe is hard to say, but we should expect some differences from year to year in the speed riders can attain given their power on the day.

Or put another way, any power estimates from ascension rates comparing year to year will have some error depending on how the localised wind plays out. The climb obvious has many changes of direction, and wind at rider level is different to the prevailing conditions (normally measured at 10m above ground level and as a rough estimate it's about half that at rider level). Of course localised wind will be shaped by the Alpe itself as well as boundary layer features such as trees, road cuttings, vehicles and so on.

The prevailing wind was from the North East in 1997, 1999, 2008, 2011 and 2015; from the North West in 2003 and 2013; from the South West in 2001 and 2006 and from the West in 2004.

Course profile shows the climb is not a constant gradient:

Fastest five ascents up Alpe d'Huez from this year's stage were:

and here are the fastest 5 riders by year (click to see larger version), with lines marking the time of the 50th and 100th fastest ascents of all time:

Read More......

Friday, July 17, 2015

Climbing power estimates: Windbags II

No specific comment, I just wanted to create a public link to the following 2014 study investigating the accuracy of climbing power estimates and to include a graphic and quote the study's conclusion.

My earlier comments on this topic of estimation accuracy can be found in this post from two years ago:

The study is:
Accuracy of Indirect Estimation of Power Output From Uphill Performance in Cycling 
Grégoire P. Millet, Cyrille Tronche, and Frédéric Grappe
International Journal of Sports Physiology and Performance, 2014, 9, 777-782 © 2014 Human Kinetics, Inc.


Study Conclusions:

Aerodynamic drag (affected by wind velocity and orientation, frontal area, drafting, and speed) is the most confounding factor. The mean estimated values are close to the power-output values measured by power meters, but the random error is between ±6% and ±10%. Moreover, at the power outputs (>400 W) produced by professional riders, this error is likely to be higher. This observation calls into question the validity of releasing individual values without reporting the range of random errors.

Read More......